Скачать или смотреть CivicGuard | Multi-Agent Crisis Intelligence | Google + Kaggle Capstone Challenge

CivicGuard is a multi-agent AI system built for the Google + Kaggle Capstone Challenge. It detects crises like floods and storms, validates evidence from multiple sources, summarizes events, and recommends safety actions — all designed to work reliably even offline.

🔗 Project Links

1. GitHub Repository : https://github.com/Devyanikhande/CivicGuard
2. Kaggle Notebook Submission: https://www.kaggle.com/code/devyanikhande/...

🔗Architecture Highlights
1. Parallel ingestion agents (social media, weather, news)
2. Sequential validation agent for evidence scoring
3. Memory engine tracking shelters & context
4. LLM-stub summarizer for offline crisis briefs
5. Risk & action engine with severity, confidence, and recommended actions

What I Learned from the Google AI Agents Intensive Course
1. Designing specialized multi-agent systems
2. Using shared memory for context coordination
3. Creating offline-safe deterministic agents
4. Adding confidence scoring for transparency
5. Building parallel → sequential agent pipelines

CivicGuard applies these concepts to demonstrate how multi-agent AI can support communities during emergencies.
CivicGuard,
multi-agent AI,
AI agents,
crisis detection system,
emergency response AI,
disaster detection,
AI crisis monitoring,
Google AI Agents Intensive,
Kaggle Capstone Challenge,
Google Kaggle capstone,
offline AI system,
LLM stub summarizer,
parallel agents,
AI validation engine,
AI safety system,
risk assessment AI,
AI architecture,
flood detection AI,
storm detection AI,
social media crisis detection,
AI for public safety,
multi-agent architecture,
AI engineering project